Can an embodied approach to social cognition accommodate mindreading, our ability to attribute mental states to another person? Prima facie it might not. Mindreading has been conceived in terms of what Susan Hurley calls the classical sandwich picture of the mind. On this view, perception corresponds to input from world to mind, action to output from mind to world, and cognition as sandwiched in between. It separates perception and action, and takes cognition to be central and distinct from both. Embodied cognition rejects the classical sandwich. The mind evolved to guide action, to enhance an organism' s coping with the world. Perception and action and action are central to cognitive activity rather than peripheral. If mindreading is a central cognitive activity, isolated from sensorimotor processing, it's hard to see how it can have a place in an embodied account of social cognition. But at the same time it seems that a complete embodied social cognition should be able to accommodate mindreading, as it is an important social cognitive skill. The present dissertation investigates how this can be done. More specifically, it investigates whether the simulation theory of mindreading can be integrated with an embodied account of social cognition. The simulation theory of mindreading holds that one understands another not by applying a theory but by using one's own mental processes to generate information about the mental processes of the other. Even though its proponents appear to have subscribed to the classical sandwich, simulative mindreading could be a central component of an embodied social cognition.